Voicea with Mohamed El-Geish: GCPPodcast Episode 162
Hi, and, welcome to episode number, 162. Of the weekly Google cloud platform podcast. I'm, mark Mandel and I'm very excited to be joined yet again by Gabby Ferrara how, are you doing today Gabby I am very good thank you Mark for asking, how are you I'm doing, okay, you're good to be joined, by you yet again on the podcast yeah I love. Being here it's nice. It's. Great so, today we are going to be talking to mohamed. El gush talking. About voice ayah and voice applications. Which I'm pretty excited by yes, then he will show us a different ways, how the approaching. Ml, to solve their problems that's, gonna be fun yeah it's really awesome and then afterwards. We have a question of the week that I think you're gonna ask me yeah so, mark, what. If I'm working in the terminal in called shell and I want to move to another computer, how. Can I continue my work that's, a great question so we'll answer that later. Okay. So why don't we get stuck into that cool thing of the week I think you had something great up first so we announced this week you will be able to query inside bigquery. Without having a credit card honest, so that means that. People like students, and other types of organization. Can learn. How to use V query without, having, to provide financial, information and, it provides up to 100 by the month of querying, capacity, which is a lot for, a free tier so that's, exciting for me that's, very very cool awesome, and. My cool thing of the week we has a great article talking about exploring, container security, encrypting, kubernetes secrets with cloud KMS, this, is actually really really cool so if you've used kubernetes, kubernetes. Has a thing called secrets, where you can store data in it especially things like passwords and whatnot but by, default the kubernetes secrets, are stored in plain text which if, you want to put like really secret stuff and there is not ideal but there is a new, thing in gke which is application, layer secret encryption that is in beta right now that you can take advantage of it's, pretty, straightforward to set up all you really need to do is specify the key that you want to use inside cloud kms to manage your secrets and pretty. Much you're good to go if you have a look at the article it shows you a little getting started guide as well as links to some, documentation to get you going cool also. This week using that and the Golden State Warriors. Announced, they are gonna be improving their fan experience. By using App, Engine, and firebase. Together. For maps so you have the blog posts there explaining, how it's gonna be done and I, have to confess mark I did not kno what Golden State Warriors work. We. Admittedly, we had to look it up we had to go headline we're like oh it's basketball in. Oakland, close, to where I live. I. Feel like that's something I should know yeah I'm sorry people that's.
All Right there's a really great article by our VP of engineering, melody McPhee Seoul who we are huge fans of and you can actually listen to her more in episode, number 158. From the podcast as well melody. Works very deeply in DevOps land and she, wrote an article talking, about a recent survey sponsored, by Google cloud by, Harvard, Business Review analytic, services talking. About DevOps and basically the road to DevOps, and in the blog post they outlined seven steps for DevOps as well as a link to the survey for more information but in there there's piloting. A small project being an open-source player embedding, security, within a software development process apply DevOps best practices, provide a massive training establish. And no blame culture build, a culture that supports DevOps, etc. Etc so check, out the blog post for all the details as well as links to the survey as well and, to. Finish, her list The Telegraph in UK announced. A partnership with, Google cloud to, help them, to manage their media, they're, gonna be using cool, things like auto ml, to classify, content, so imagine. All of those contents, that they have from way, way back. Being able to classify out of ml and be more, searchable, more easy, to access now that's, very cool and I think I think we're trying to get them on the podcast too won't we yes, they're, gonna be on the podcast and, they're gonna be able to talk more about that I'll be interviewing them so stay tuned well why don't we go have a chat with our friend Mohammed over at Boise yeah yeah let's go I am, exceptionally delighted, to be joined by Mohammed LK chief, architect, at voice yeah coming, to chat with us today on the podcast about everything they do how you doing today Muhammad, good, how are you doing. Pretty well thanks. So much for joining us today and talking to us all about voice yeah oh thanks. For having me so Mohammed what, do you do is voice eeeh and who, are you tell us all. Right so my name is Mohammed oh gosh I'm chief, architect, and co-founder, of Lycia I'm, responsible, for design. And other architecture, tasks. Here and I record I do, some other, tasks, founders, to like. Change evil light bulbs and all that so almost, everything you can imagine at startup its restore II. Feel. Like that's a thing yeah everywhere, founders always have to do that's awesome so tell. Us about voice II air like at a high level what is this company do yeah so as you can imagine from the name or all about voice and enterprise, specifically. So. If you think really about like how we use, voice in our, day-to-day especially. At work they're, like almost three big categories one. Of them is actions, that, you take so imagine for example if you have like an assistant that collects.
Our Google assistant. And then you talk to the assistant and execute, the actions for you like you know turn on the lights and so on the. Big category. For us is that workflow, integrations, that you actually want to automate, a lot of actions, and enterprise, you. Can imagine, you walk to a meeting and you can a lost you don't know like what the agenda is what is gonna happen after. The meeting is done and we want to automate the process of distilling other outcomes, of the meeting in a, very concise. Way and converting. Those into actions, so you can say like our assistant. Is Eva what's the stock price of Google today create a task for me to follow up with IT regarding. Security policies, and this becomes an actual action, or tasks, in your task management system like turtle or you know your system of choice and then, the other big categories information. So you are in a meeting and you want to capture, or query some information, from another system so like for example like, you can ask if okay Eva so, networks. Or social, media I'm not doing. Is. Focusing, on the conversations, I'm focusing on the people there so that's. The biggest value add I think when, you're in a meeting you're very focused you're present, and you have really high attention, a queue so, attention. Code since I think so that's the water was look for and you. Want to be focused on the conversation, and the people around you and Evo will basically take care of like taking notes one, you to assistant. Like this is an important moment and this is an action I want to be following. Up with that sounds, really cool you've also got integrations. With a bunch of different systems you mentioned like action items and actually, doing things can you tell us what does that actually look like in my day to day you might like sending notifications, to by email or what. Are the cool things I can do right, so a lot of these systems actually allow you to like send.
Them Emails. To, create, something in them you, can, for example create a card and turtle or create, a task, in JIRA, or a ticket and we, plug-in into a lot of them so you can actually go to our website and explore with other systems, we will, plug into and you, can figure it easily with a single like email address and if. I will send anything that you want to capture into, these systems and basically, starting, that workflow, kicking off that, action, which is maybe creating a note or creating a task and so on and so forth if he met like you have your notepad, and your pin and you walk to a meeting you can actually talk and you can tell Eva that this is important, and this is what I wanted to capture instead. Of like scribbling and you, know maybe losing, a few minutes here and there of that meeting you, could have spent focusing, more on the people around you at the conversation, for sure awesome, so yeah, you're the chief architect, and this is the GC podcast, like when I get into the technology side of things what, technologies, do you use to make a reality I'm guessing, there's kind of two parts to it I know there's like probably the transcription, but then there's also like pulling out the highlights and the action items and actually doing the actions so what are the bits and pieces that you're using here right, so we're. Really big on kubernetes, it's, something that I have, vidacup, appreciation. For coming, from a background where I actually build. Something very similar like you're soft it, was really hard it was really complex, and we understand, like how life. Right now for startups, and companies that, are growing is much easier than it used to be like a decade ago so, you can actually just walk into, your terminal and type, in a few commands and voila, you have a cluster running in GCP which is amazing, and you don't have to worry I've heard a lot of things like provisioning, and scale and so on and so forth so this, is one of the biggest items we have that, we kind, of don't. Worry about so, we, let Google manage the kubernetes, installation, for us and we worry about the application, and growing the business. And basically the infrastructure, is kind, of well taking you know care of and in the gke world which is the google kubernetes engine so our biggest. Like focus, now is just how to understand. The. Business and the growing the integrations, and growing like you know the machine learning systems, we have and we, don't worry about like auto scaling or these things of course we understand how Jiki works and we know how to tweak it you know like our systems are very, kind of complex, as well but it takes all of complexity, away for sure so, compute engine jakie stackdriver, all. Of these tools like you know also, that storage systems, and whatnot are really, great for any startup to like go. To GCP, and like start working on growing, their business and not worrying about managing. You know services, like that so with, all of this tooling, that you use sometimes. You face, up challenges. Because, you, have a learning curve you, need to learn how to stock driver I need to learn how to use days how, to use that how. This affect. The development, and the. Challenges, that you ended up facing, for. Your product, and voice yeah that's a good question so there, are layers of abstractions. That, might. Isolate you, from the. Underlying, problems, that a lot of distributed, systems face so, back in the day when you're building your own like, set up your own cluster, you had to worry about like where, I'm gonna you know give my machines, so, you know if I have to buy or provision hardware I have to put it in a data center I actually have to worry about that rack space I have to worry about networking, ever to worry about security physical security after. About, a lot of things and then after all this you still have to worry about the, actual. Systems, that are running the. Databases. And file, storage and backing.
Up Systems, redundancy, high availability and, on, top of all of this your application, has to be up and running and available and performance, and all these nice. Characteristics. So, if you think about the balance, between doing, all of the above versus learning a framework or learning something that's I think, it really thoroughly simpler, like, how to work with kubernetes, or how to work with stackdriver I think we're in a much much, better, place today it of, course can improve, meaning that there actually some. And some other companies whose. Entire business model is basically simplifying, this even more, but, I think it's the right balance for what. I would call like understanding. The the complexity, underneath or basically peeling the layers of the onion if you're, working on an operating system you need understand a little bit of the foundations. Of this operating. System how, it schedules, tasks, how, it manages, files and, so, on and so forth like in the, world. We live in you have understand, the data. Center operating system which is in our case kubernetes. Like how it schedules like you know docker containers, and what. Happens when you, know some conditions are met and you know you might. Be in a risk of losing your data or not so, there of course some constructs you have to understand, but for the most part I think we're in a much much, much better space or, place than, before and that. Allows, us to move faster, and focus on our own application. Concerns, versus, like you know learning, way deeper. Things in the stack and of, course you, know the the thing I will emphasize in this is the more you know the, better you are like knowledge is power and like if you understand kubernetes very. Well and if you understand how that stack driver works you're, gonna be able to like perform, faster, and move faster yes I'd, love to hear more about like the flow of information - in this system because I'm guessing there's there's more to it than just Krupa Nettie's right I'm guessing there's like some audio coming in there I'm processing maybe and then are you running custom ml models you putting like other tools in it like what's the what's, that flow of information how's, that how's that all getting processed yeah it's a great question so our basic. Input is audio, right and there's some metadata that are. Associated. With such audio recordings, so we join a meeting and we record what's happening, we also respond, to, commands and that's technology, in, machine learning known, as the keyword spotting, so, we don't this in a very like real-time, fashion. That you want to respond right, away to the users, commands, and execute. Them so, there's a really good like separation, here between what. We can. Classify as, the real-time versus. Like Nearline versus. Offline, processing. Of audio. And other metadata, we would consume and we. Depend heavily on like binary stacks, the schedule. All this workload. For, us so, the flow of the data is basically the audio is, being captured real-time, and we're responding to commands using machine learning algorithms, in legal time and we, do one pass on the transcript, and then we capture, the audio and store, it and then we do another multi, passes, on that and we, produce the transcript. For, the for the user at the end so as you can see that you know the. The biggest thick focus, for us is actually how to be always available capture.
The Audio and not drop it and also, always be performant, with really low latency, to, respond to these commands so that when you're having, this experience which is very fairly unique you have a, very. Powerful computer, or personal, assistant on the phone with you on the conference line so, we're talking now on this line and we can actually invite. Eva and either we'll be capturing. This and responding, to my commands, right I'm telling Eva like highlight this or this is an important note or take an action and as. An action item and so on and so forth so all. Of these have to be very performant. And have very low latency so. We, understand, the, complexities. Of like, how, to route, audio and RTP. Or real-time, protocol. On top, of communities, and we. Have like certain, requirements. Around, like, CPU, utilization and memory utilization for, these. Systems. That are processing, audio in real time which. Is very important to understand that you know the constraints, of the system when you're using one, scheduler, versus the other and we're, really, happy with the band's we have today in terms of like how to like separate. These like workloads in different node pools and the how to schedule them auto-scaling, so that we are not always like you know over, provisioning, the machine so our, workload. Varies, according to like you know business, hours in different locations, so we can actually kind of get, an idea of when we're gonna need more machines versus we're gonna need fewer, machines and and so, on and so forth for each system so, it's a very churning problem but, kubernetes, is helping us manage, that very well yeah, that's fun what, are you building your ml models in oh so, that's a great question about. Machine, learning in general and, like the philosophy, or taking. And, building. Systems, so, as a very fresh, or early startup when we started, a couple of years ago we. Had this like list, of principles about, software, engineering in general and specifically, about like AI, and, machine learning in which, we set a set, of guiding principles for. Basically. Coming, up with these decision-making framework so, when you ask somebody in the team like hey what, are you gonna use to build this model there should be a very easy kind. Of list of criteria to like make a decision, so for example like as a company, that's centered. Around voice and the voice of like our users is very important to us to protect security. Is always top priority for us so, a lot of these systems like. In machine learning already, get about like you know like security. Like we're serving them through, kubernetes so, the training part of machine learning happens, of course like a flying but, if you're serving something online we're protecting. The data or making sure that everything, is secure, encrypted. Then, we talk about like other frameworks, characteristics. That meet, the other set of priorities for us or guiding principles, so you talk about bit, of iteration, and speed of iteration, or velocity. Of iteration, is, extremely. Important in machine learning machine. Learning by definition or, the way it's it's done it's, very iterative, it's. Extremely, elusive meaning that you start with some idea and you, build some hypotheses.
And Then you say this, will work on paper and you actually go and deploy it and the wild and you get some feedback and you say like okay I'm gonna have to improve this and now you collect. More data and you keep iterating on it so there. Are a lot of frameworks today like tensorflow and others make, this process, extremely easy we have also some flexibility, around what. Framework, you use thanks. To a new framework called onyx that allows, you to build models using multiple, frameworks. And at the end you can serve them using. One of them or a single, like engine, that you choose and you can serve from that so we have some flexibility and freedom that we allow folks on the team to actually use, multiple. Frameworks including. On the cloud if they wish, but at the end of the day we try to standardize the serving mechanism, so that you always serve on communities, yes, so. With, all this machine learning that you do if, models, how. Do you see, you. Serving. The enterprise, world where, everyone. Has a different demand, everyone. Wants a piece of you in a different way how. Do you see, serving, enterprise, and world yeah so there's this concept. Of. Well-posed. Machine, learning problems, in the, sense that these kind, of problems when. You're trying to solve them you, build, up something and you, put it out there and users. Have, some experiences. With them and then, they use the data that. Come from the users or the results of their output, as feedback, that, they can learn more and this, is where we are actually it's it's a virtuous, data network, effect meaning, that if you are using voice here today and let's, say you're in a meeting and then if, it transcribes something or it's, a task. For you and there's, a very, like, jargony. Word that you use that if I did not recognize you can go to our interface, you can fix it there and then we, will learn from that so that the next time we would actually you know make our best to like mitigate, that or or fix it for you and then the more users we have from the same industry, we, can have a more, explicit. Effort to, actually prepare. That the words that are pertaining. To that industry, or that domain. First, before, you actually use the system or as, you, can see the data effect, the more you use it the more it learns from you and then, becomes even, more, accurate. So, that's really a good place to be at or your problems, will post ooh but. What happens if you have the same term being, used across different industries in. Different ways yeah. That's when actually, you have to collect some metadata about hosts talking, and it's more personalized, and then you can disambiguate. Between the two how cool so, they might sound the same but, they actually might be spelled. Differently, but because you know who's talking first, like you know I know you're in your email address I know some data about the speaker, so, it's more personalized, towards you so you basically need the contextual, information right, right and we get that because you actually you, know sign up with us we, know who you are and we, kind of let. Evil learn from, your. Indecent capacities basically, so, that's how you learn in accents, too because, I have problems of some assistance, because when I say some. Words they. Don't get it what I'm saying and accent. Is really hard you know yeah, that's part of the solution yeah nice, that's, pretty cool so obviously, you're doing a lot of voice. Stuff, is, this just the beginning do, you plan on doing more where do you see the voice interaction experience. Kind of growing I would, really love if I could have demo our system, because we are also introducing. Some video elements. Specifically. When it's a presentation, or slides that your song inside the context of meeting it's, actually very interesting, to see the interaction, between the voice and the playback of the slides and the, highlights of the meeting the minutes, that we captured, it's very rich experience and, we're also looking for. Like something that we can capture, whether, it's audio or video to.
Get More context, about the meeting and give it to you in a consumable. Format, that, is easy, to like you know integrate. With your workflow so, today we actually you know also will have some playback, of the slides or the video ad what, is the most interesting, voice. Application, you created. At voice see the, most like. Challenging. And interesting. Thing, you did with that technology, I have. To think a little bit about this there are a lot of challenges in finding a start-up for sketch yeah definitely. When, it deals with like, non-deterministic. Input. Like speeds so. You have a lot of machine, learning domains, that are easier, than than, speech or perceptual, you know knowledge. Acquisition so. For example if you're thinking, of like recommendation, systems which is something I used to do at my belief is life at LinkedIn I think, you have more leeway. Or you have more freedom. And Nick showing the user like recommendations. And so on unlike for example the newsfeed but when you're dealing with like human. Perceptual, like, tasks, like I've heard something and I can make sure that you know I need to make sure that this is the transcript for it it's, a bit more challenging I want to say that for sure one, of the most challenging ones is keyword spotting, and you know responding, to commands it's, almost a magical moment when. You walk to a meeting and then can. Talk to the conference, line and you, get stuff done that's for sure a very like, challenging. Process. And to changing, problem because you, need to understand the intent of what the users saying you, need to understand, where, this information should flow and you need to convert that data, into, an actionable, outcome. The combination. Of these problems, I think is the most challenging and most interesting for sure fantastic. What's been the on the customer side that what's been the most interesting integration, you've seen all, definitely integrations, with task management systems, because a lot of people like, to keep track of what they're doing and they think of meetings as a, place where they get work done and, a place where they can coordinate and, discuss, ideas and, then, they come out with them some like maybe handwritten, notes or maybe, some cryptic, notes they captured. During, the meeting cuz you all really want to be engaged and they just like type a few words here and there but then now, they can just get, other experience. With like very nicely. You know fished, out lists. Of outcomes, that they. Just entered using our, input using their voice I think, this is actually where voice. Shines. It's, the natural way of, communicating. With computers, so anybody is interested in, human-computer. Interaction. Or HCI. We, can tell you that in the beginning we had the terminal we had like just, words and a command, line and, very thick you know dull, way of interfacing because, that was the limitation, of that era and, then we advanced. A little bit to the graphical, user interface, or GUI and, also. Because like you know we couldn't do better than that if you maybe saw, some of the early systems, that used voice or other natural, UI, methods. Of input, they, weren't that good back, in the day so we had to like you know resort, to like the mouse or the planner and then, some icons. And whatnot but, now with the richness, of, machine, learning techniques, and deep learning and all the advances, we had in the past I would say five to six years it, became possible to actually walk to a room and have, a $20 device that talks, to you and you, can do things with it which is in my, opinion like it's, just a revolution, of the UI, and, the way we interact with computers overall. Does. That mean II should have been an email you probably, are answering. That question if we see it right now right because. All, the meetings you'll, get all the stuff keywords, what, is important, you, sent to the integration. Systems, for task management so. In the end you. Have everything. That you need, just. By having your device, or Eva help. You I think, I think it's a very fair point, in the future, I wouldn't. Exclude us from doing something similar to tell you like maybe you shouldn't attend this meeting. As. A matter of fact we were working on a lot. Of things that will help you make the decision one. Of it is that we have as humans. Intrinsic, fear of missing out or formal, but. You don't actually have to sit, there for an hour and listen to the meeting if you can send Eva on, your behalf and if, I come back reporting, to you like here are the important minutes you, can just you know spend. One minute or a couple minutes to listen to like the important minutes, or read them and worked, on this very, amazing concept, in collaboration, which, is teams. And channels so, let's say I manage. A team and we have a meeting with a client and, they, want to share this meeting with me so we, can just create a channel and then I have access to the channel that has.
All This like clients, meetings and then I can coach them through it or I can follow. Up with the the customers, if I if I needed to be in that meeting and I couldn't make it for example or I'm double booked or triple booked so for most very important, and we want to specify, that. It by, giving you like here's what's really happening. With, this excuse, of meetings in a very Sesenta way like here, the highlights and I think you know you can get. A really good high value from. Just, reading, those or following. Up on those instead of attending, all of these meetings and I, wouldn't be surprised if we, end up answering. The questions should this meeting has be, any minute said mm-hmm, all, right or is it gonna be like don't have meetings with John John tends to waffle that. Sounds. Great, if people want to learn more about voice EO like where should they go to learn about this stuff so. Definitely, dub-dub-dub, we, see a dot-com. Also on LinkedIn, and Twitter Google. Voice eeeh and interact. With us I would love to hear your feedback and, sign, up today it's. Free. And get a trial for our premium solution, and I, hope you enjoy it cool and if anyone wants to learn about any of the stuff you've talked about today it sounds like you're working on some really interesting problems do you only resources that you recommend yes so definitely. Feel, free to reach out to us we. Can get, you a demo, and if you want each other personally I can, also add my. Contact information, here and anybody, on the team we. Have an about page and contact, us page, on voice datacom fantastic. Before, you run away is there any other things, that we haven't managed to touch on that you want to make sure that people know about I think, as a parting. Note this, is a very exciting era and I think folks, like, we've. Heard a lot of things about AI. Taking, over the world and AI being, this fourth, revolution after, the industry revolution, how it's gonna, change the shape. And form of how we do work I want, to emphasize the way where we're, doing things and thinking, about AI, more. Of an augmented, intelligence, so it's not here to replace your job some jobs you know over time will evolve and change and, some, of them are gonna disappear but, I would send. A message to everybody to say like think, of AI is your friend and make, sure that you know. More about like what's happening, because it's gonna affect how we do work in the future its, accelerated, definitely, on an accelerated path right now of. Course it's here to improve your job not to take care of a different, AI thank, you Mohamed awesome, all right well Mohamed thank you so much for hanging out with us today and chatting to us all about voice yeah thank you so much thanks for having me thank, you mohamed al guys for, helping us understand. More what the various CIA is doing and all, the insights that you gave on machine, learning on this episode yeah, super cool episode thank you so much for joining us so mark the. Moment that you're waiting for yeah, question, the week what, if I'm working a terminal called shell and I want to move to another computer, how, can I continue my work so, that's a really good question and I found this out the other day I was really excited so if, you haven't use cloud shell it's kind of Awesome it's, basically an integrated, terminal that's built into cloud console I also has a matching and a whole bunch of other fun things so it means that if you just want to like get something done with g-cloud or maybe cutesy TL or something like that and a bunch of other tools that come with it as well it's just a really handy way of bringing up a terminal and getting things done what's. Super cool about it is not only do you get like persistent dis storage for a certain amount of time as well so you can leave stuff in there and if you move from computer to computer as long, as you're accessing that on a regular basis it stays there which is really great what, I didn't realize and the, implications, of which is it, actually uses a piece of software called T MUX in the background if you haven't used that locally it's basically a terminal, session management it keeps those those terminals alive and persistent, so, what happens is say, you're working on one computer and you're like that was cool and then you go to the office and you're working on your workstation for example and you're like oh I wonder if there's still a running process running or I want to check what's going on you, open up your terminal and cloud, shell will be like oh hey I transferred, your processes, over from, like the old computer to this one so, you can still see what's coming on and it's all the same so for example I was doing a bunch of cleanup I had a slew.
Of App engine versions, lying around, and so I'd written a script to just delete a whole bunch it was gonna take about half an hour and. I basically started. It at home and then got on a bus and gone to work got, to work and then fired up cloud shell on my workstation which, wasn't my laptop and it was like I wonder if it's finished yet and I popped it up and I could see it was still processing and I was in a completely different browser in a completely different computer and I was like that is awesome, mind blowing there's. This until. Joke that, clock, shows the best product, that we don't advertise. So. There's, only on the podcast pose, with the feature so take a look they're absolutely, fantastic. All right Gabby what, are you doing what have you been up to are you going anywhere cool so, I'm going, to the Museum, of Natural History I never been there before but. I'll be giving a workshop on, MLA. PI's and called functions, for their Brown Scholars, Program, which it is forever cool, high school girls so, that's, gonna be nice because it's, gonna be in Python, and they know more Python, than I do probably so. Okay so you'll teach each other things yeah. That. Sounds good that sounds great fantastic, and you Mark what you're gonna be doing what am I gonna be doing I think the the thing that's coming up most in March game developers, conference we. Will be there all, of Google will be there in all its glory if you're in - cloud stuff though make sure to check out all the sponsor sessions that we're running on Wednesday we've basically blocked, out the day so if you wanna learn about it Kahn is our up and match our machine learning or all sorts of other cool stuff that we're applying to games definitely swing, by and then there's cloud next that is coming two weeks after, that I assume Gabby you will be there as well at cloud next with me yes. Excellent, excellent so I'll be presenting there as well and, I'm sure you'll be there doing other fun stuff as well I hope, so. Excellent. And then, way later down the line in April, I'll be at the East Coast games conference as well so, lots of fun things lots of fun things and travel, to do yeah exactly, I probably won't be streaming this week just because I need to get on a plane and go back to Australia for a bit but otherwise I'm sure I'll hop on there at some point again soon okay, I hope you solve your problems. There. Are always more problems to solve their never-ending list. Of problems okay, right okay thank you all right Gabi thank you very much for joining me this week again, and thank, you everyone for listening and we'll see you all next week, see. You all later.